cuda: pull in libdir when linking C/C++
* In `CudaDependency._detect_language`, the first detected language is considered the linking one. Since `nvcc`/`cuda` implicitly know where the cuda dependency lives, this leads to situations where `cpp` as linking language is erroneously detected as `cuda` and then misses the `-L` argument.
This commit is contained in:
parent
8cb16b2d6a
commit
33adc420f3
|
@ -24,7 +24,7 @@ if T.TYPE_CHECKING:
|
|||
|
||||
class CudaDependency(SystemDependency):
|
||||
|
||||
supported_languages = ['cuda', 'cpp', 'c'] # see also _default_language
|
||||
supported_languages = ['cpp', 'c', 'cuda'] # see also _default_language
|
||||
|
||||
def __init__(self, environment: 'Environment', kwargs: T.Dict[str, T.Any]) -> None:
|
||||
compilers = environment.coredata.compilers[self.get_for_machine_from_kwargs(kwargs)]
|
||||
|
|
Loading…
Reference in New Issue